Counts of cases and deaths are key metrics of COVID-19 prevalence and burden, and are the basis for model-based estimates and predictions of these statistics. I present here graphs showing these metrics over time in Washington state and a few other USA locations of interest to me. I update the graphs and this write-up weekly. Previous versions are here.
See below for caveats and details. I originally posted updates on Mondays but have switched to Wednesdays to accommodate the current Washington DOH data release schedule.
Figures 1a-d show case counts per million for several Washington and non-Washington locations. The Washington locations are the entire state, the Seattle area where I live, and the adjacent counties to the north and south (Snohomish and Pierce, resp.). The non-Washington locations are Ann Arbor, Boston, San Diego, and Washington DC.
Figures 1a-b (the top row) show smoothed data (see details below); Figures 1c-d (the bottom row) overlay raw data onto the smoothed. The figures use data from Johns Hopkins Center for Systems Science and Engineering (JHU), described below. When comparing the Washington and non-Washington graphs, please note the difference in y-scale: the highest current Washington rate (about 1100 per million in Pierce) is about the same as Ann Arbor and well below Boston (about 1800 per million).
The smoothed graphs for Washington (Figure 1a) show that rates continue to decline and are far below their recent peaks; less happily, the rates remain above the first two waves in Spring and Summer 2020. The raw data ((Figure 1c) and simple trend analysis (described below) indicates that the most recent data (4 weeks) is too variable to be confident in the direction, but looking back 6 or 8 weeks, the decline is clear. The graphs for non-Washington locations (Figures 1b,d) are falling dramatically. Trend analysis concurs, although for Washington DC, we have to look back 6 weeks to get a strong signal.
Figures 2a-d show deaths per million for the same locations. When comparing the Washington and non-Washington graphs, again please note the difference in y-scale: the current Washington rates (11-27 per million) are well above Ann Arbor (3 per million) and similar to the other non-Washington locations (23-31 per million).
The smoothed Washington data (Figure 2a) shows three waves. The second peak was thankfully lower than the first; the third wave exceeded the first in all areas except Seattle (King County). The graphs are well down from their recent peaks; the rates are continuing down in the state as a whole and Snohomish (north of Seattle) but seem to be leveling off in Seattle and Pierce (south of Seattle). The raw data (Figure 2c) remains quite variable; trend analysis (described below) indicates that the most recent data (4-6 weeks) is too variable to be confident, but looking back 8 weeks, the decline is clear.
The smoothed non-Washington data (Figure 2b) shows early peaks in most locations, followed by a long trough, followed by a second wave starting in November. The graphs are well down from their recent peaks and seem to be continuing down. The raw data, though variable, broadly supports this view. Trend analysis is variable also but generally supports a downward direction.
The next graphs show the Washington results broken down by age. This data is from Washington State Department of Health (DOH) weekly downloads, described below. An important caveat is that the DOH download systematically undercounts events in recent weeks due to manual curation. I extrapolate data for late time points as discussed below.
Figures 3a-d are cases. The graphs are split into 20-year age ranges starting with 0-19, with a final group for 80+.
Early on, the pandemic struck older age groups most heavily. Over time, cases spread into all age groups, even the young. During the second wave, older groups did better in most locations with young adults (20-39 years) becoming the most affected group. The third wave swept into all age groups with young and middle aged adults (20-39 and 40-59 years) leading the surge. As the wave has grown, the oldest people (80+) are again strongly affected. The curves are heading down in all age groups, but caution is in order because the data remains variable.
Figures 4a-d are deaths. These graphs aggregate 0-59 into a single group, since the death rate in these ages is near 0.
The shocking devastation of the 80+ age group jumps off the page. Statewide (Figure 4a), the death rate in the 80+ group shows three waves. The early death rate for this group in Seattle (King County) (Figure 4b) reached over 600 per million reflecting the early outbreak at a long term care facility in the area. Deaths in Snohomish (north of Seattle) (Figure 4c) for the 80+ group peaked early in the pandemic, then declined and stayed fairly low but, in the third wave, climbed considerably above the early Seattle peak; some of the increase reflects an outbreak in a long term care facility in the county. Deaths in Pierce (south of Seattle) (Figure 4d) for this group stayed fairly low until mid-November but then surged almost to the level of the early Seattle peak. The decline at the latest time points in all locations may be due to reporting lags or other data problems, including the variability shown in Figure 2c.
The term case means a person with a detected COVID infection. Until the recent reporting change, Washington DOH data limited this to “confirmed cases”, meaning people with positive molecular COVID tests, but going forward they plan to separate out “probable cases”. Other states already do this, but the data source I use here only includes “confirmed cases” (or so I believe based on the name of the file I download).
Detected cases undercount actual cases by an unknown amount. As testing volume increases over time, it’s reasonable to expect the detected count to get closer to the actual count. Some of the increase in cases we see in the data is due to this artifact. Modelers attempt to correct for this. I don’t include any such corrections here.
The same issues apply to deaths to a lesser extent, except perhaps early in the pandemic.
The geographic granularity in the underlying data is state or county. I refer to locations by city names reasoning that readers are more likely to know “Seattle” or “Ann Arbor” than “King” or “Washtenaw”.
The date granularity in the graphs is weekly. The underlying JHU data is daily; I sum the data by week before graphing.
I truncate the data to the last full week prior to the week reported here. Thus, data for the March 3 update includes counts through the last week of February, ending February 27.
I smooth the graphs using a smoothing spline (R’s smooth.spline) for visual appeal. This is especially important for the deaths graphs where the counts are so low that unsmoothed week-to-week variation makes the graphs hard to read. In versions of the document prior to December 30, 2020, I used a 3-week rolling mean for this purpose.
The Washington DOH data (used in Figures 3 and 4 to show counts broken down by age) systematically undercounts events in recent weeks due to manual curation. I attempt to correct this undercount through a linear extrapolation function (using R’s lm). I have tweaked the extrapolation repeatedly, even turning it off for a few weeks. The current version uses a model that combines date and recentness effects.
The trend analysis mentioned above computes a linear regression (using R’s lm) over the most recent four, six, or eight weeks of data and reports the computed slope and the p-value for the slope. In essence, this compares the trend to the null hypothesis that the true counts are constant and the observed points are randomly selected from a normal distribution. After looking at trend results across the entire time series, I determined that p-values below 0.1 indicate convincing trends; this cutoff is arbitrary, of course.
DOH provides three COVID data streams.
Washington Disease Reporting System (WDRS) provides daily “hot off the presses” results for use by public health officials, health care providers, and qualified researchers. It is not available to the general public, including yours truly.
COVID-19 Data Dashboard provides a web graphical user interface to summary data from WDRS for the general public. (At least, I think the data is from WDRS - they don’t actually say).
Weekly data downloads (available from the Data Dashboard web page) of data curated by DOH staff. The curation corrects errors in the daily feed, such as, duplicate reports, multiple test results for the same incident (e.g., initial and confirmation tests for the same individual), incorrect reporting dates, incorrect county assignments (e.g., when an individual crosses county lines to get tested).
The weekly downloads lag behind the daily feed causing data for the last few weeks to be incomplete. I attempt to correct this undercount through a linear extrapolation function (described [above]{#techdetails}).
The weekly DOH download reports data by age group: 20-year ranges starting with 0-19, with a final group for 80+.
The DOH download includes data on hospital admissions in addition to cases and deaths, although I don’t show this data here.
In past, DOH updated the weekly data on Sundays, but as of December 22, 2020 they switched to Mondays. When Monday is a holiday, they release data on Tuesdays.
JHU CSSE has created an impressive portal for COVID data and analysis. They provide their data to the public through a GitHub repository. The data I use is from the csse_covid_19_data/csse_covid_19_time_series directory: time_series_covid19_confirmed_US.csv for cases and time_series_covid19_deaths_US.csv for deaths.
JHU updates the data daily. I download the data the same day as the DOH data (now Tuesdays) for operational convenience.
I use two other COVID data sources in my project although not in this document.
New York Times COVID Repository. The file I download is us-counties.csv. Like Washington DOH and JHU, NYT has county-level data. Unlike these, it includes “probable” as well as “confirmed” cases and deaths; I see no way to separate the two categories.
COVID Tracking Project. This project reports a wide range of interesting statistics (negative test counts, for example), but I only use the case and death data. It does not provide county-level data so is not useful for the non-Washington locations I show. The file I download is https://covidtracking.com/data/download/washington-history.csv. I use this only as a check on the state-level Washington data from the other sources.
The population data used for the per capita calculations is from Census Reporter. The file connecting Census Reporter geoids to counties is the Census Bureau Gazetteer.
Comments Please!
Please post comments on Twitter or Facebook, or contact me by email natg@shore.net.